Facial Expression Notebook

1. Code for affectnet-yolo-format dataset downloading in colab or working with it directly in kaggle
2. PyTorch model creation, training, validation and testing

ONNX Notebook

1. Export ONNX model from the pytorch model
2. onnex model validation and testing

Deployment

1. The main.py file has the FastAPi script for the local deployment
2. make sure to update the images folder and the model paths in the file
3. Install FactAPi: pip install fastapi uvicorn
4. run: uvicorn main:app --reload

Dataset link

https://www.kaggle.com/datasets/fatihkgg/affectnet-yolo-format

Models Link

PyTorch models: https://drive.google.com/drive/folders/12-tpSJAc-tq1bDYKwN5tfhYQOdIUVzuX?usp=drive_link

ONNX models: https://drive.google.com/drive/folders/128PDDXSYdEv0ytusLmlcfwD60Hi02Je-?usp=drive_link

Project documentation

The "Facial Expression Detection AI Model" file provides a comparison of different models.

Demo: Local deployment with the onnx model of > 0.78 mAP

Facial.Recognition.mp4